Re: PIDs vs flow IDs

[alain.ritoux@6wind.com]

Marie-Jose Montpetit wrote:
> I'm working on a new version of the AR draft and something has come up: 
> how PIDs are not really flow IDs and can change. As far as I know in 
> broadcast TV networks when you use INT with PID, a (re)MUX may re-mark 
> the PID, and then could remark the
> INT. This is a standard MUX function. However in a data network I think 
> it is assumed that PIDs are to be used end-to-end (as unidirectional 
> flows), and Muxes should not re-mark. If that is the case it's ok. If 
> it's not the case then we may have to find a way to easily identify 
> flows and map them appropriately.

My understanding is that  PID should really be end-to-end, because
what MUX seem to be able to do is a sort of NAT-like behaviour : it
changes the packets, AND it performs some ALG (i.e. by re-writing INT).

If we think about some possible dynamic mapping (sort of ARP/NDP)
using whatever return link may be available (SCPC, UDLR, ...) then
the MUX re-writing of PID will be catastrophic ! andI think a quite
dynamic mechanism would be REALLY usefull, for static tables such as
INT may not be enough for "automatically" numbered networks (with some
Prefix Delegation in IPv6, ...).


One other question : the Adress Resolution should not only provide
(addr) --> (mac, PID) mapping, but ALSO some precision about the
encapsulation used, i.e. (PID) --> (method) mapping to be able to swtch
between MPE and ULE for exemple.

Regards.
Alain.
--
Alain RITOUX
Tel +33-1-39-30-92-32
Fax +33-1-39-30-92-11
visit our web http://www.6wind.com